# starting by only exploring the CY* variables
endowment_data <- read_rds("./data/endowment_filter_data_990.RDS") %>%
  select(EIN, fiscal_year, contains("CY"))

companies_to_ein <- read_csv("./data/companies.csv") %>%
  mutate(EIN = as.character(ein)) %>%
  select(EIN, organization_name)

# only include EINs that have at least one observation of 
# one of the endowment variables
include_eins <- endowment_data %>%
  pivot_longer(-c(EIN,fiscal_year)) %>%
  group_by(EIN) %>%
  summarize(na_count = sum(is.na(value)),
             total_rows = n()) %>%
  filter(na_count < total_rows) %>%
  pull(EIN) %>% unique()


# taking most recently available information from the CY* variables
# this is to take into account findings from cross referencing that 
# values are not always in correspondence
# when they are in correspondence, this will be equivalent to just taking
# the CY data from each fiscal year 
endowment_data <- endowment_data %>%
  filter(EIN %in% include_eins) %>% 
    group_by(EIN) %>%
    pivot_longer(3: ncol(.),
                 names_to = "variable_name") %>%
    mutate(source = ifelse(grepl("CYM", variable_name),
                           substr(variable_name, 1,4), "CY"),
           # get the year lag from the CYM variable name
           # CYM1 corresponds to 1 year lag
           year_lag = ifelse(grepl("CYM", variable_name), 
                             substr(source, 4,4), 0),
           year_lag = as.numeric(year_lag),
           fiscal_year = as.numeric(paste0(fiscal_year)),
           # get rid of CY or CMX part of the variable name (where X is a year)
           variable_name = gsub("CY|CYM.", "", variable_name)) %>%
    mutate(value_year = fiscal_year -year_lag
           ) %>%
  group_by(EIN, value_year, variable_name) %>%
  arrange(EIN, variable_name, fiscal_year) %>%
  # pick the most recent one available
  slice_max(n = 1, order_by = fiscal_year)  %>%
  select(EIN, value_year, variable_name, source, value) %>%
  rename(fiscal_year=value_year) %>%
  ungroup() %>%
  # add company names
  left_join(companies_to_ein)

# this function splits the data into 4 groups based on the mean value for that EIN 
# and plots the variable over time, faceted by group
# this is to handle the fact that the values are on very different scales,
# so if we plot them all together, it masks any interesting trends
# dotted lines represent where there was a missing year between observations
plot_variable <- function(var) {
  
  # get eins with at least one observation of the variable
  eins_with_variable <- endowment_data %>%
  filter(variable_name == var) %>%
    group_by(EIN) %>%
    summarize(number_observations = sum(!is.na(value))) %>%
    filter(number_observations != 0) %>%
    pull(EIN)
  
  # split into quantiles based on EIN mean
  data <- endowment_data %>%
      filter(EIN %in% eins_with_variable & variable_name == var) %>%
      group_by(EIN) %>%
      mutate(EIN_mean = mean(value,na.rm= TRUE)) %>%
      ungroup() %>%
      group_by(variable_name) %>%
      mutate(quantile_group = ntile(EIN_mean, n = 4)) %>%
      group_by(EIN) %>% 
      # make sure EIN has single quantile group
      mutate(quantile_group = max(quantile_group)) %>%
      mutate(quantile_group_labels = factor(paste0("Quantile ", quantile_group))) %>%
      mutate(organization_name = ifelse(is.na(organization_name),
                                        "Not Available", 
                                        organization_name)) %>%
    ungroup()
  
  # data with no nas so we can connect the values wehre there was a missing year
  # instead of just having no line connecting points from years on each side of the gap
  data_no_nas <- data %>% filter(!is.na(value))
  
  data %>%
    ggplot(aes(x = fiscal_year, y = value, color = EIN, label = organization_name)) +
    geom_line(data = data_no_nas,
              aes( 
                  x = fiscal_year,
                  y = value, group = EIN),
              color = "darkgray",
              linetype = "dotted") +
    geom_point(size = .9) +
    geom_line() +
    facet_wrap(~fct_reorder(
      quantile_group_labels,
      .x = quantile_group), scales="free_y", ncol = 2) +
    scale_y_continuous(labels = comma) +
    scale_x_continuous(breaks = seq.int(2010, 2021, by =2)) +
    viridis::scale_color_viridis(option = "mako", discrete = TRUE, end=.94) +
    theme_bw() +
    labs(title = paste0("Change in ", var, " Over Time"),
         x = "Fiscal Year",
         y = paste0("value of ", var)) +
    theme(plot.title = element_text(size = 18, 
                                    hjust = .5, face="bold",
                                    margin =margin(.1,.1,10,.1)),
          plot.subtitle = element_text(hjust = .5, face="italic"),
          axis.text.x = element_text(size = 13),
          axis.title = element_text(size = 13, face = "bold"),
          legend.position = "none")
}